Lessons learned from running Claude Code swarms at scale

Fleet is a devtool that orchestrates swarms of coding agents (Claude, agy, Codex) via a centralized Python controller, recently adding a UI for easier management. It’s designed for developers experimenting with multi-agent coding workflows at scale, offering insights into coordination, failure modes, and throughput. The project is interesting because it tackles the practical challenges of running concurrent AI coders, revealing lessons about agent reliability and system design that go beyond single-agent use.